Introduction

Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is a major analytical tool for the characterization of proteins [1]. An important advantage of MALDI is that intact peptide ions can be produced without abundant fragmentation, which facilitates rapid characterization of a protein digest. Furthermore, the use of specific reducing matrices such as 2,5-dihydroxybenzoic acid [2], 2-aminobenzoic acid [3], and 1,5-diaminonaphthalene (1,5-DAN) [4] can cause in-source decay (ISD) during MALDI experiments. In particular, the most impressive of these matrices is 1,5-DAN, which efficiently produces fragment ions and allows rapid characterization of the sequence of intact proteins [5]. MALDI-ISD is initiated by laser-induced hydrogen radical transfer from the matrix to the analyte peptide [6, 7], which has been suggested to occur via hydrogen bonding between the peptide and the matrix prior to desorption [8]. The resulting aminoketyl radical intermediate eventually undergoes N–Cα bond cleavage, leading to the formation of a c′/z• fragment pair [9, 10]. Subsequently, the radical z• fragment either reacts with the matrix or undergoes further fragmentation to give various fragments such as, z′, z, w, and [z + matrix] [11, 12]. Because z′ fragments often appear as intense peaks, protein sequencing by MALDI-ISD is performed by interpreting mass differences between series of consecutive c′ and z′ ions. To obtain accurate mass differences, MALDI-ISD employing a Fourier transform ion cyclotron resonance (FTICR) mass spectrometer was recently developed [13,14,15]. Additionally, a MALDI-ISD–based pseudo-MS3 method, that is mass selection of ISD ions followed by fragmentation with post-source decay, provides information on the identity of the C- and N-terminus of the protein and enables direct identification of a target protein in a mixture [16,17,18,19]. Moreover, MALDI-ISD preferentially produces c′ and z′ fragments without degradation of labile post-translational modifications, so the locations of phosphorylation [20], O-glycosylation [21], and polyethylene glycosylation [22] in proteins are determined. Consequently, MALDI-ISD is becoming increasingly important in the field of proteomics, and this is likely to continue.

In contrast to the N–Cα bond cleavage, the cleavage of the Cα–C bond was recently found to occur when 5-nitrosalicylic acid (5-NSA) was used as an oxidizing matrix for MALDI-ISD [23]. MALDI-ISD with 5-NSA permits identification of phosphorylated and isoaspartate residues in peptides [24, 25]. More recently, hydroxy-nitrobenzoic acid isomers [26,27,28], 7,7,8,8-tetracyanoquinodimethane derivatives [29], and 4-nitro-1-naphthol (4,1-NNL) [30] were reported to efficiently produce fragment ions due to a Cα–C bond cleavage. Regarding the mechanism, hydrogen abstraction from the peptide by the matrix produces a hydrogen-deficient peptide radical intermediate, which undergoes Cα–C bond cleavage. Because hydrogen abstraction can occur at all residues with similar probability, the Cα–C bond cleavage by MALDI-ISD is less prone to specific bond cleavages. However, as exceptions, the presence of proline (Pro), sarcosine (Sar), and glycine (Gly) residues was reported to affect the Cα–C bond cleavage efficiency [31]. We found that the Cα–C bond cleavage on the N-terminal side of the Pro and Sar residues did not occur; instead, fragments due to a peptide bond cleavage were observed [31]. The lack of fragments due to the Cα–C bond cleavage is likely to arise from the lack of amide hydrogen on the peptide backbone of the Pro and Sar residues. Consequently, the Cα–C bond cleavage by MALDI-ISD with an oxidizing matrix is assumed to be initiated by abstraction of an amide hydrogen from the peptide. Upon formation of the nitrogen-centered peptide radical, the cleavage of Cα–C bonds located on the N- and C-terminal sides of the radical site provides a•/x and a/x• fragment pairs, respectively, and the corresponding fragmentation pathways are illustrated in Scheme 1a, b, respectively. As described above, the presence of the Pro and Sar residues suppresses the cleavage of Cα–C bonds at the N-terminal side of Pro and Sar residues, because these residues do not contain an amide hydrogen on the peptide backbone. These results indicate that the resulting nitrogen-centered radical undergoes cleavage of Cα–C bonds located on the N-terminal side of the radical site (Scheme 1a) and the Cα–C bond at the N-terminal side of Pro and Sar residues are not cleaved because the Pro and Sar residues cannot form the nitrogen-centered radical [31]. In addition, the density functional theory (DFT) calculation also suggests that the Cα–C bonds located on the N-terminal side of the radical site in nitrogen-centered radicals are preferentially cleaved [32]. Therefore, I concluded that MALDI-ISD using an oxidizing matrix produces an a•/x fragment pair through a nitrogen-centered radical intermediate (Scheme 1a) [33, 34].

Scheme 1
scheme 1

Proposed mechanism of Cα–C bond cleavage by MALDI-ISD. (a) a•/x fragment pair formation from a nitrogen-centered radical by the cleavage of Cα–C bonds located on the N-terminal side of the radical site. (b) a(N=C)/x• fragment pair formation from a nitrogen-centered radical by the cleavage of Cα–C bonds located on the C-terminal side of the radical site. (c) a(C=C)/x• fragment pair formation from a β-carbon–centered radical

Full size image

The pathway described in Scheme 1a is considered to be induced by the fragmentation techniques involving electron ejection from multiple deprotonated peptides, such as electron detachment dissociation (EDD) [35], negative electron transfer dissociation (NETD) [36]. These techniques result in the formation of a charge-reduced peptide anion that contains a radical site on the carboxyl group of the side chain or C-terminal carboxyl group. Because the carboxyl group often interacts with hydrogen at the peptide bond, the resultant radical at the carboxyl group abstracts a hydrogen from the backbone amide nitrogen, leading to an a•/x fragment pair through a nitrogen-centered radical intermediate [37]. Therefore, MALDI-ISD using an oxidizing matrix would share some mechanistic similarities with EDD and NETD.

In contrast, the presence of the Gly residue affects the Cα–C bond cleavage efficiency, as in the case of the Pro residue [34]. Recently, Nagoshi et al. [30] focused on the low cleavage efficiency of Cα–C bonds at the C-terminal side of Gly residue. These authors focused that the low fragmentation efficiency arises from the lack of a β-carbon atom in the Gly residue, and they suggested the MALDI-ISD process involves a β-carbon–centered radical intermediate (Scheme 1c). The site of hydrogen abstraction from peptides was also investigated by these authors using a peptide with deuterium labeling at the side chain of the alanine residue, Ala-(d3). MALDI-ISD of the deuterium labelled peptide employing an oxidizing matrix produced a fragments containing two deuterium atoms, indicating that the formation of a ions involved deuterium abstraction from the side chain of the Ala-(d3) residue. Because they assumed that MALDI-ISD involves Cα–C bond cleavage leading to a/x• fragment pair formation (Scheme 2b, c), intense signal of a-d2 fragments originated from β-carbon–centered radical intermediates. According to this hypothesis, the generation of a β-carbon–centered radical intermediate is more favorable than the generation of a nitrogen-centered radical intermediate during the MALDI-ISD process. However, the authors did not consider the contribution of a•/x fragment pair formation through a nitrogen-centered radical (Scheme 1a), which, is concluded as the most probable pathway by my review articles [33, 34], as described above. Subsequently, the resulted a• radical immediately undergoes hydrogen atom abstraction by reaction with the matrix, and potential sites for this hydrogen abstraction are the amide nitrogen and β-carbon atoms (Scheme 2a). Consequently, a-d2 fragments can be produced from nitrogen-centered radical intermediates by a•/x fragment pair formation (Scheme 1a) and subsequent deuterium abstraction from β-carbon atoms in a•-d3 radicals (Scheme 2a).

Scheme 2
scheme 2

Formation mechanism of a(C=C) and a(N=C) fragments originating from the cleavage of the Cα–C bond at the C-terminal side of Ala-(d3) residue

Full size image

To estimate the contribution of each of the pathways described in Scheme 1 to the MALDI-ISD process, this article focuses first on DFT and transition state theory (TST) calculations that can be used to predict the mechanisms underlying the Cα–C bond cleavage, followed by a discussion of experimental and theoretical results. Subsequently, the formation mechanism of d fragments, which are often observed in MALDI-ISD mass spectra [26,27,28,29] is investigated. Such d fragments are believed to originate from a• radicals, as shown in Scheme 3 [26, 29], and their yield is found to be dependent on the transition state barrier to the side-chain loss of a• radicals. The joint experimental and computational study suggested that a•/x fragment pair formation through a nitrogen-centered radical intermediate (Scheme 1a) is the most probable pathway for MALDI-ISD employing an oxidizing matrix.

Scheme 3
scheme 3

Formation of d fragment from a• radical by radical-induced cleavage

Full size image

Experimental

Materials

The peptide, substance P, was purchased from the Peptide Institute (Osaka, Japan), and all synthetic peptides were purchased from GenScript (NJ, USA). Table 1 shows detailed peptide information. 5-NSA and 4,1-NNL were purchased from Tokyo Chemical Industry (Tokyo, Japan). Ammonium bicarbonate, iodoacetamide, 3-hydroxy-2-nitrobenzoic acid (3H2NBA), and 3-hydroxy-4-nitrobenzoic acid (3H4NBA) were purchased from Wako Pure Chemical Industries (Osaka, Japan). All reagents were used without further purification. All solvents used were of HPLC grade quality, except for water, which was purified using a Milli-Q purification system (Millipore, Billerica, MA, USA). The conductivity of the purified water was 18.2 MΩ/cm.

Table 1 Monoisotopic mass (Mm), amino acid sequence, and composition of analyte peptides used
Full size table

MALDI-ISD Experiments

The peptides were dissolved in water to give a 50 μM solution. DCL-substance P was S-carbamidomethylated by 0.22 M iodoacetamide in a 100 μL solution of 0.1 M ammonium bicarbonate at pH 8.0 for 30 min at room temperature. The resultant mixture was removed by GL-Tip SDB (GL Sciences, Tokyo), and then, the eluted peptide solution was dried by centrifugal concentration. The matrices 5-NSA, 3H4NBA, 3H2NBA, and 4,1-NNL were used for the MALDI-ISD experiments. The matrix was dissolved in a 1:1 (v/v) water/acetonitrile mixture to give a 10 mg/mL solution. Then, 0.5 μL of the peptide aqueous solution (50 μM) and 0.5 μL of the matrix solution were deposited onto the MALDI plate. MALDI-ISD mass spectra were recorded using an AXIMA-CFR plus (Shimadzu/Kratos, UK). The analyzer was operated in linear mode with pulsed ion extraction. The laser power was optimized to obtain MALDI-ISD mass spectra that had high signal-to-noise ratios for the ISD ion peaks. In total, 500 shots were accumulated for each MALDI-ISD mass spectrum.

Calculations

All electron structure calculations were performed using the Gaussian 16 program [38]. The geometries of the radicals and molecules were obtained by optimization based on DFT calculations using the PW6B95D3 [39], which is recognized the best conventional global hybrid functional[40]. The structures were characterized by frequency calculations at local energy minima at the PW6B95D3/6-31+G(d,p) level. To establish the energetics for fragmentation, transition state geometries were also optimized at the PW6B95D3/6-31+G(d,p) level and confirmed by checking whether their vibrational frequencies had an imaginary part. The connections between the transition states and the reactants and intermediates were checked by intrinsic reaction coordinate analysis [40] starting from the transition state geometry. The harmonic frequencies obtained by frequency analysis were used to obtain zero-point energy corrections. The rate constants were calculated by the TST with one-dimensional semiclassical tunneling correction [41] assuming the asymmetric Eckart potential by using the Gaussian Post Processor program [42].

In the present study, Zubarev’s notation was adopted for peptide fragment ions [43]. According to this notation, homolytic Cα–C bond cleavage yields the radical a• and x• fragments, and abstraction of a hydrogen atom from the a• or x• radical by matrix produces an a or x fragment, respectively.

Results and Discussion

Mechanism of Cα–C Bond Cleavage

To examine the processes occurring during MALDI-ISD that arise from hydrogen-deficient peptide radicals, the mechanisms underlying the Cα–C bond cleavage were investigated by DFT calculations. Because MALDI-ISD induces a similar fragmentation in both positive and negative ion modes, the MALDI-ISD process is considered as a charge remote fragmentation [34]. Consequently, first, I focused on the fragmentation process of a neutral peptide, and the effect of charge on the fragmentation process is then discussed.

To examine the fragmentation process, the hexapeptide-acetylated Ala-Ala-Ala-Ala-Ala-Ala-amide (AcA6-NH2) was used as a model system for the DFT calculations at the PW6B95D3/6-31+G(d,p) level. First, I searched for the model peptide conformation with the maximum number of intramolecular hydrogen bonds to find the global minimum-energy structure, because the peptide in the gas phase is stabilized by intramolecular hydrogen bonding. The polyalanine is known to adopt a conformation corresponding to the α-helix in the gas phase [44]. As expected, the DFT calculations yielded a minimum-energy conformation for AcA6-NH2 corresponding to the α-helix motif, which was then used as the initial conformation for the calculation of the fragmentation pathway. Then, I considered the cleavage of the Cα–C bond at Ala3-Ala4 to determine the most probable fragmentation pathway during MALDI-ISD; three competitive fragmentation pathways have been proposed, as shown in Scheme 1. According to the proposed processes, I calculated the fragmentation pathway for the Cα–C bond cleavage at Ala3-Ala4 in AcA6-NH2, as summarized in Figure 1.

Figure 1
figure 1

Cleavage mechanism of the Cα–C bond at Ala3-Ala4 in the hydrogen-deficient radical form of AcA6-NH2. Scheme 1 describes the reaction pathways. The relative energy levels (kJ mol−1) were obtained from DFT calculations at the PW6B95D3/6-31+G(d,p) level and include zero-point vibrational energys

Full size image

Regarding the fragmentation pathway in Scheme 1a, hydrogen abstraction from the amide nitrogen at the Ala4 residue in AcA6-NH2 induced Cα–C bond cleavage at Ala3-Ala4. Consequently, [AcA6-NH2−H]• with a radical site on the amide nitrogen at Ala4 (R1a) was used as the initial conformation. As expected from the proposed mechanism in Scheme 1a, R1a produced an a3•/x3 fragment pair by the Cα–C bond cleavage on the N-terminal side of the radical site. To obtain the transition state conformation during the Cα–C bond cleavage, the corresponding bond length was increased until an energy maximum was reached, and then, the obtained conformation was further optimized to find the saddle point. The calculated fragmentation pathway is shown in Figure 1a. The nitrogen-centered radical R1a undergoes Cα–C bond cleavage through the transition state TS1a, and the corresponding energy barrier was 27 kJ mol−1. As a result, the a3• and x3 fragments produced were linked together via inter-fragment hydrogen bonding to provide the complex IM1a, which is more stable than R1a by 25 kJ mol−1. The products a3• and x3 are 57 kJ mol−1 less stable than R1a.

In contrast, [AcA6-NH2−H]• with a radical site on the amide nitrogen at Ala3, R1b, undergoes cleavage at the Cα–C bond between the Ala3 and Ala4 residues by the fragmentation pathway shown in Scheme 1b. This fragmentation pathway was also examined by DFT calculations (Figure 1b). As expected from Scheme 1b, the cleavage of the Cα–C bond on the C-terminal side of the radical site produces an a3(N=C)/x3• fragment pair through the transition state TS1b, which is 50 kJ mol−1 higher in energy than the reactant, R1b. The product a3(N=C) has a double bond between the amide nitrogen and the α-carbon, and the counterpart x3• radical contains a hydrogen-rich isocyanate radical in which the radical is delocalized over the isocyanate (N=C=O) group. The resultant a3(N=C) and x3• are linked together via inter-fragment hydrogen bonding, producing the complex IM1b, and the energy of the a3(N=C) and x3• formation was higher than that of R1b by 120 kJ mol−1.

Upon forming the nitrogen-centered radical, the cleavage of the Cα–C bonds located on the N-terminal and C-terminal sides of the radical site occurs competitively, as shown in Scheme 1a, b. To determine the most probable fragmentation pathway during MALDI-ISD, I have, therefore, calculated the rate constants for the dissociation of R1a and R1b (Figure 2). As Figure 2 shows, the rate constant, k, is dependent on the temperature of the peptide radical. Because the temperature of the peptide ion is estimated to be around 900 K and 450 K in the early and late stages of MALDI plume, respectively [45], MALDI-ISD occurs at below 900 K. Regarding the timescale of the fragmentation, MALDI-ISD occurs in the ion source before applying the acceleration voltage, which is set to several hundred nanoseconds after laser irradiation. In consequence, MALDI-ISD occurs at a 100-ns timescale, indicating that a dissociation pathway with a rate constant of greater than 107 s−1 occurs during MALDI-ISD. According to Figure 2, the formation of an a3•/x3 fragment pair from R1a is faster than the formation of an a3(N=C)/x3• fragment pair from R1a at below 900 K, as shown in Figure 2. Therefore, the calculation strongly suggests that the nitrogen-centered radical attacks the Cα–C bond located on the N-terminal side of the radical site to provide an a•/x fragment pair, as indicated by the fragmentation pathway shown in Scheme 1a and Figure 1a. To confirm the dissociation rate of R1a and R2a, the DFT calculation is also performed with the M06-2X/6-31+G(d,p) level and the calculated rate constants are shown in the Supplemental Information, Figure S1. As in the case of Figure 2, Figure S1 indicates that the nitrogen-centered radical preferentially produced an a•/x fragment pair through the pathway shown in Scheme 1a. This result agrees with the conclusion drawn from the experimental results summarized in my previous review articles [33, 34], and the fragmentation process producing an a3(N=C)/x3• fragment pair through a nitrogen-centered radical intermediate (Scheme 1b) can be ruled out during the MALDI-ISD process.

Figure 2
figure 2

Rate constants for Cα–C bond dissociations in the hydrogen-deficient peptide radicals R1a, R1b, and R1c. The horizontal dashed line at k = 107 corresponds to dissociations occurring on the MALDI-ISD timescale of 100 ns

Full size image

In contrast, Nagoshi et al. [30] proposed a Cα–C bond cleavage process involving a β-carbon–centered radical intermediate (Scheme 1c), as described in the Introduction Section. Herein, the β-carbon–centered radical at the radical site of the Ala3 residue, R1c, was used as the initial structure for the calculation of the a3(C=C)/x3• fragment pair formation. As Figure 1c shows, R1c produced the a3(C=C)/x3• fragment pair by Cα–C bond cleavage, and the corresponding transition state was TS1c, which is 90 kJ mol−1 higher in energy than the reactant, R1c. The resultant complex comprising a3(C=C) and x3•, IM1c, is less stable by 55 kJ mol−1, and the complete dissociation to a3(C=C) and x3• fragments requires 125 kJ mol−1. The rate constant for the corresponding fragmentation process is also plotted in Figures 2 and S1, which indicates that the dissociation of R1c through TS1c can occur on the timescale of the MALDI-ISD experiment (100 ns). Consequently, the a3(C=C)/x3• fragment pair can be generated if the β-carbon–centered radical is produced during MALDI-ISD.

The effect of charge on the MALDI-ISD process was then investigated. For this purpose, a hexapeptide containing a basic residue, acetylated Lys-Ala-Ala-Ala-Ala-Ala-amide (AcKA5-NH2), was used as a model system. The [AcKA5-NH2+H]+ containing a protonated Lys residue, which is the most preferential site for protonation, was used as the initial structure in DFT calculations. As for AcA6-NH2 (Figure 1), the Cα–C bond cleavage at Ala3-Ala4 in [AcKA5-NH2+H]+ was examined by DFT calculations, and the results are summarized in the Supplemental Information, Figure S2. By comparing Figures 1 and S2, it can be seen that the transition state barriers obtained are similar, indicating that the presence of the extra proton does not influence the MALDI-ISD process. Consequently, MALDI-ISD can be viewed as a charge remote fragmentation, which further suggests that, in general, Figure 1 can be applied to both positive- and negative-ion MALDI-ISD processes. In summary, nitrogen-centered and β-carbon–centered radicals produced a3•/x3 and a3(C=C)/x3• fragment pairs, respectively, under MALDI-ISD experimental conditions, and the fragmentation producing an a3(N=C)/x3• fragment pair (Scheme 1b) is unfavorable during the MALDI-ISD process. Consequently, I then focused on the reaction of the a3• radical with 5-NSA to investigate the production of a3(N=C) and a3(C=C) fragments.

Mechanism of a (N=C) and a (C=C) Fragment Formation

To determine the structure of the a fragment, Nagoshi et al. [30] investigated a peptide, RLGALGACLADLAEL, containing deuterium-labeled Ala residues, Ala(d3), at positions 10 and 13. Both a-d2 and a-d3 fragments were produced from the deuterium-labeled peptide by MALDI-ISD using 4,1-NNL as the matrix [30]. Similarly, we have reported previously the MALDI-ISD mass spectrum of the deuterium-labeled peptide, RLGNQWA(d3)VG(d2)DLAE, when 5-NSA was used as the matrix [23]. In the previous article, we focused on the a9 ion originating from a Cα–C bond cleavage at Gly(d2)9–Asp10, and the results indicated that the formation of a ions did not involve hydrogen abstraction from the α-carbon [23]. Figure S3 in the Supplemental Information shows the MALDI-ISD mass spectrum of the deuterium-labeled peptide (the original data were previously published as Figure 5b in reference [23]). In the present work, I focused on Cα–C bond cleavage at Ala(d3)7–Val8 residues, leading to the formation of the a7 ion. As Figure S3 shows, MALDI-ISD of the deuterium-labeled peptide produced both a7-d2 and a7-d3, which are identified as a(C=C) and a(N=C) fragments, respectively. Although the abundance ratio of a ions containing Ala(d2) and Ala(d3) residues would be dependent on the matrix used and the sequence of the analyte peptide, the experimental results clearly indicate that MALDI-ISD employing an oxidizing matrix produces both a(C=C) and a(N=C) fragments (Scheme 2). Moreover, the MALDI-ISD mass spectrum clearly shows the a fragment without deuterium abstraction, a-d3, which is not formed through a β-carbon–centered radical intermediate (Scheme 2). Consequently, the presence of an a-d3 fragment can be viewed as evidence for the contribution of nitrogen-centered radicals in the MALDI-ISD process.

To assess the suitability of Scheme 2a, the formation of a(N=C) and a(C=C) by the reaction between a• radicals and 5-NSA was investigated using DFT calculations. For the purpose of these calculations, a3(AcAAA)•, which is the product in Figure 1a, was used as a model. Figure 3 summarizes the calculated reaction pathways for the formation of a(N=C) and a(C=C) fragments. With regard to the formation of a(N=C), the interaction between the a3(AcAAA)• and 5-NSA produced the complex R3a, and the corresponding binding energy was 58 kJ mol−1. Hydrogen transfer from the amide nitrogen in a3(AcAAA)• to the nitro group in 5-NSA produced an a3(N=C) fragment through the transition state TS3a, which is 12 kJ mol−1 higher in energy than R3a.

Figure 3
figure 3

Mechanism of the formation of a3(N=C) and a3(C=C) by the reaction between a(AcAAA)• radicals and 5-NSA. Scheme 2a describes the reaction pathways. The relative energy levels (kJ mol−1) were obtained from DFT calculations at the PW6B95D3/6-31+G(d,p) level and include zero-point vibrational energys

Full size image

In contrast to R3a, the most stable complex consisting the a3(AcAAA)• and 5-NSA is R3b, which is 96 kJ mol−1 more stable than the reactants. An a3(C=C) fragment is formed by the hydrogen abstraction from the β-carbon in a3(AcAAA)• through the transition state TS3b-1, and the corresponding barrier is 44 kJ mol−1. Although TS3b-1 is more stable than TS3a, reaction barrier for the formation of a3(C=C) is higher than a3(N=C). In contrast, R3b-2 is another conformation of the complex, in which a3(AcAAA)• and 5-NSA are linked together via a hydrogen bond between the amide hydrogen in a3(AcAAA)• and the carboxyl oxygen in 5-NSA. R3b-2 is 63 kJ mol−1 less stable than R3b-1. R3b-2 produced a3(C=C) through the transition state TS3b-2, and the corresponding barrier is 16 kJ mol−1. Although the transition state barrier for the hydrogen transfer is dependent on the conformation of the complex consisting the a3(AcAAA)• and 5-NSA, the transition state for the formation of both a3(N=C) and a3(C=C) is more stable than that of the reactants (free a3(AcAAA)• and 5-NSA). In consequence, the formation of the a(N=C) and a(C=C) fragments can be considered to proceed as a barrierless reaction pathway, the reaction between a3(AcAAA)• and 5-NSA would give both a(N=C) and a(C=C) fragments.

In summary, the results for a deuterium-labeled peptide indicate that MALDI-ISD produced two different types of a fragments, i.e., a(N=C) and a(C=C). Since both types of a fragment can be produced from nitrogen-centered radicals by the Cα–C bond cleavage and subsequent hydrogen abstraction, the fragmentation pathways shown in Scheme 1a, c cannot be discriminated by MALDI-ISD experiments with peptides containing a deuterium-labeled Ala residue.

Mechanism of d Fragment Formation

To discriminate between the fragmentation pathways shown in Scheme 1a, c, next, I focused on the d fragment, which is formed by the side-chain loss of an a• radical, as shown in Scheme 3, because a(C=C) cannot produce d fragments. MALDI-ISD using an oxidizing matrix was reported previously to induce a side-chain loss at Asp, Leu, and Ile residues [26,27,28,29]. In this study, substance P–related peptides were used as models to investigate the yield of d fragments, which can reflect the amount of a• radical produced by MALDI-ISD. Figure 4 shows the MALDI-ISD mass spectra of substance P, NEI-substance P, and DCL-substance P. Since substance P–related peptides contain basic Arg and Lys residues at positions 1 and 3, respectively, MALDI-ISD preferentially produced protonated a and d fragments. As Figure 4 shows, the side-chain loss of the a• radical occurred at the Gln, Glu, Asn, Asp, Leu, Ile, and Cys. In particular, intense signals of the d7 and d8 fragments were observed in the MALDI-ISD mass spectra of DCL-substance P instead of the a7 and a8 fragments (Figure 4c). This result indicates that MALDI-ISD effectively induces the side-chain loss of Asp and Cys residues. In contrast, d fragments originating from the Cα–C bond cleavage and subsequent side-chain loss of Phe, i.e., d7 and d8 of substance P, d6 of NEI-substance P, and DCL-substance P, were absent.

Figure 4
figure 4

MALDI-ISD mass spectra of (a) substance P, (b) NEI-substance P, and (c) DCL-substance P. 5-NSA was used as the matrix

Full size image

To understand the formation mechanism of the d fragments, I calculated the dissociation rate of an a• radical with the sequence Ac-Ala-Ala-Xxx, where Xxx is Asp, Asn, Glu, Gln, Leu, Ile, Phe, or Cys. Based on the fragmentation mechanism in Scheme 3, the energy of the transition state and products of the dissociation were calculated (Table 2). The transition state barriers for the d3 fragment formation were around 95–130 kJ mol−1, except for a• radicals with the sequences Ac-Ala-Ala-Cys and Ac-Ala-Ala-Phe. The side-chain loss of the Cys residue required only 70 kJ mol−1, and the fragmentation proceeded without transition state. In contrast, the transition state barrier for the side-chain loss of the Phe residue was 163 kJ mol−1, which is higher than that of the other residues. In consequence, a• radicals originating from a cleavage on the C-terminal side of the Cys residue preferentially give d fragments compared with other residues. Indeed, MALDI-ISD mass spectrum of DCL-substance P shows an intense signal of d8 fragment, which is formed by the cleavage of Cys-Leu bonds, compared with a8 fragments (Figure 4c). The use of a different matrix, 4,1-NNL, is reported to show an intense signal of d fragments originating from the Cα–C bond cleavage on the C-terminal side of the Cys residue, and the abundance of corresponding a fragments is low [30].

Table 2 Energy profiles for the side-chain loss of a3• with an Ac-Ala-Ala-Xxx sequence, where Xxx is Cys, Glu, Gln, Asp, Asn, Leu, Ile, or Cys*
Full size table

Regarding the side-chain loss from a• radicals, I found that S-carbamidomethylation on the Cys residue decreases the required energy for the fragmentation. The transition state barrier for the side-chain loss of a• radical with the sequence Ac-Ala-Ala-Cys*, where Cys* indicates a carbamidomethyl Cys residue, was only 34 kJ mol−1 (Table 2). To estimate the yield of d fragments, the rate constants for the side-chain loss of a3• radicals with an Ac-Ala-Ala-Xxx sequence, where Xxx is Cys*, Phe, Asp, Asn, Glu, Gln, Leu, or Ile, were calculated (Figure 5). Notably, the dissociation of a3•(AcAAC) proceeds without a transition state barrier and the corresponding rate constant could not be estimated. Regarding the a3•(AcAAC*), the dissociation constant is ~ 109 s−1 at 450 K and ~ 1011 s−1 at 900 K, indicating that the corresponding dissociation occurs within 1 ns, even in late stages of MALDI plume. Because a• radicals originating from a cleavage on the C-terminal side of the Cys* residue immediately undergo side-chain loss, the S-carbamidomethylation of DC*L-substance P would enhance d8 fragment formation and suppress a8 fragment formation. Indeed, MALDI-ISD of DC*L-substance P with 5-NSA shows an intense signal of d8 fragment, which can be understood by the high dissociation rate constant (Figure 6a). Subsequently, I have analyzed DC*L-substance P with recently reported matrices: 3H2NBA [28] and 4,1-NNL [30] (Figure 6b, c), to determine the influence of the matrix used for the yield of d fragments. As Figure 6 shows, the side-chain loss of the Cys* residue was efficiently induced and the d8 fragment was observed as an intense signal for all tested matrices. In contrast, the abundance of a8 fragment was very low. Because the d fragments originated from the a• radicals, the yield of d8 fragments in the MALDI-ISD of DC*L-substance P reflects the amount of a•8 radicals produced. In contrast, a(C=C) fragments produced through a β-carbon–centered radical intermediate could be estimated from the yield of a8 fragments in the MALDI-ISD of DC*L-substance P, because the a(C=C) fragment does not produce d fragments. Indeed, the d8 fragment is 10–20 times more abundant than the a8 fragment in the MALDI-ISD mass spectra of DC*L-substance P (Figure 6). The intense signal of d8 fragments originating from the cleavage of the Cα–C bond on the C-terminal side of Cys* residue indicates that fragmentation leading to a•/x fragment pairs preferentially occurs (Scheme 1a), which suggests further that, in general, Scheme 1a holds for Cα–C bonds that do not involve a Cys* residue.

Figure 5
figure 5

Rate constants for side-chain dissociations of a3• with an Ac-Ala-Ala-Xxx sequence, where Xxx is Cys*, Phe, Asp, Asn, Glu, Gln, Leu, or Ile. The horizontal dashed line at k = 108 corresponds to dissociations occurring on the MALDI-ISD timescale of 10 ns

Full size image
Figure 6
figure 6

MALDI-ISD mass spectra of DC*L-substance P using different matrices. (a) 5-NSA. (b) 3H2NBA. (c) 4,1-NNL

Full size image

Next, I focused on the reaction rates for the side-chain loss of other residues. As described above, the timescale and temperature of MALDI-ISD are estimated to be 100 ns and 900 K, respectively. Consequently, the fragmentations with rate constants greater than 107 s−1 at 900 K occur during MALDI-ISD. As Figure 5 shows, the rate constant for the side-chain loss of a3• is around 107 s−1 at 900 K, except for Phe and Cys*. Therefore, the side-chain loss from these residues is kinetically feasible under MALDI-ISD conditions. As expected, both a and d fragments were produced by the Cα–C bond cleavage on the C-terminal side of Gln, Glu, Asn, Asp, Leu, and Ile residues. In contrast, the side-chain loss of the Phe residue did not occur during MALDI-ISD for all tested matrices, because the fragmentation is predicted to be too slow at 900 K. Consequently, the absence of d fragments originating from Cα–C bond cleavage on the C-terminal side of the Phe residue can be understood by the low dissociation rate constant.

Then, I focused on the formation of d fragments originating from the cleavage on the C-terminal side of Asp residue, which are observed as intense signals in MALDI-ISD mass spectra of DCL-substance P and DC*L-substance P (d7 fragment in Figures 4c and 6). As Figure 6 shows, the abundance of d7 fragments originating from the side-chain loss of the Asp residue is dependent on the matrix used, which agrees with the previous report [26]. In contrast, MALDI-ISD of NEI-substance P produced medium or low ion yields of d7 fragments, which originate from the Cα–C bond cleavage on the C-terminal side of Asn (Figure 4b). According to Figure 5, the dissociation rate constants of a3•(AcAAD) and a3•(AcAAN) are similar. Consequently, the intense signals arising from d fragments originating from the Cα–C bond cleavage on the C-terminal side of Asp cannot be explained by the unimolecular dissociation of a• radicals, and the side-chain loss of the Asp residue must occur through an alternative dissociation pathway. The proposed pathway is shown in Scheme 4, in which the hydrogen transfer from the carboxyl group on the Asp residue to 5-NSA potentially produced d fragments. To investigate the suitability of the proposed fragmentation pathway, DFT calculations were performed. As Figure 7 shows, the nitro group in 5-NSA binds with the carboxyl group on the Asp residue in a3•(AcAAD), forming R7, and the corresponding binding energy is 41 kJ mol−1. The hydrogen transfer from the carboxyl group on the Asp residue to 5-NSA proceeds through the transition state TS7, which is 6 kJ mol−1 more stable than the reactants (a3•(AcAAD) and 5-NSA). As expected from Scheme 4, the hydrogen abstraction from a3•(AcAAD) induces side-chain loss, leading to d3 and CO2. The products d3, CO2, and [5-NSA+H]• are 1 kJ mol−1 less stable than the reactants. The hydrogen abstraction–induced side-chain loss (Scheme 4) proceeded through a lower energetic pathway compared with the unimolecular side-chain loss (Scheme 3). Therefore, the intense signal arising from d fragments originating from the Cα–C bond cleavage on the C-terminal side of Asp can be explained by the reaction illustrated in Scheme 4.

Scheme 4
scheme 4

Proposed mechanism of the side-chain loss of the Asp residue by hydrogen abstraction from the carboxyl group in the Asp residue to the nitro group in 5-NSA

Full size image
Figure 7
figure 7

Calculated mechanism of the side-chain loss of the Asp residue by the reaction between the a(AcAAD)• radical and 5-NSA. Scheme 4 describes the reaction pathways. The relative energy levels (kJ mol−1) were obtained from DFT calculations at the PW6B95D3/6-31+G(d,p) level and include zero-point vibrational energys

Full size image

Similarly, hydrogen abstraction from the side chain of the Asn residue in a3•(AcAAN) radicals by 5-NSA would produce d3 fragments and isocyanic acid (HNCO). However, the corresponding transition state was not obtained and the reaction is calculated to be exothermic by 50 kJ mol−1. The hydrogen abstraction–mediated side-chain loss does not occur at the Asn residue due to the low stability of isocyanic acid. Both the experimental and theoretical results suggest that abstraction-induced side-chain loss only occurs for the Asp residue in a• radical.

Regarding the yield of a and d fragments, the d7 fragment was more abundant than the a7 fragment in the MALDI-ISD mass spectra of DCL-substance P and DC*L-substance P when 5-NSA was used as a matrix (Figures 4c and 6a). Similar results were obtained when 7,7,8,8-tetracyanoquinodimethane derivatives were used as the matrix [29]. Since d fragments originate from a• radicals, these experimental results strongly suggest that MALDI-ISD preferentially produces a•/x fragment pairs through the fragmentation pathway shown in Scheme 1a.

C-Terminal Side Fragments

In this section, C-terminal side fragments produced by MALDI-ISD are discussed. Figure S4 in the Supplemental Information shows the MALDI-ISD mass spectrum of the peptide PVKVYPNGAEDESAEAFR, and the original data were published previously as Figure 2d in reference 23. Since the peptide contains a basic Arg residue at the C-terminus, the C-terminal side fragments were selectively observed in the MALDI-ISD mass spectrum. As expected from the fragmentation pathway shown in Scheme 1a, x fragments were observed as the C-terminal side fragments. In contrast, x• radicals, which are formed through Scheme 1c, would undergo further radical reactions due to their low stability.

Regarding the fragmentation of β-carbon–centered radical, the ultraviolet photodissociation (UVPD) of peptide with highly labile radical precursors is reported to produce the β-carbon–centered radical, which eventually undergoes Cα–C bond cleavage, leading to a fragments [46, 47]. In contrast, the counterpart x• radicals are unstable species which are never observed in the UVPD spectrum and z• radicals are observed instead of the x• radicals [47]. The results suggest that the x• radicals are stabilized by a loss of isocyanic acid, yielding z• radicals (Scheme 5). To assess the suitability of Scheme 5 for the MALDI-ISD process, the rate constants of the corresponding dissociation processes were calculated using an x3• radical with a sequence of Xxx-Ala-Ala-NH2, where Xxx is Gly, Ala, or Asn, as a model (Figure 8). The detailed fragmentation processes calculated by DFT are shown in the Supplemental Information, Figure S5. According to Figure 8, the rate constant for isocyanic acid loss from the x• radical is ~ 109 s−1 at 900 K, indicating that it is a kinetically feasible process under MALDI-ISD conditions. The z• radical is considered to be an intermediate of MALDI-ISD with a reducing matrix, and it produced z, z′, and w fragments by radical reactions in MALDI plume [34]. Consequently, the x• radical would be converted to z, z′, and w fragments through a z• radical intermediate. With regard to the MALDI-ISD of the peptide PVKVYPNGAEDESAEAFR (Figure S4), the z•10 and z•11 radicals do not produce w fragments due to the lack of side chain in Gly and Ala residues. Consequently, x•10 and x•11 radicals produced the either z or z′ fragments. However, these fragments were absent from the spectrum shown in Figure S4, implying the absence of x• radicals during MALDI-ISD. These results also support the proposed mechanism of Cα–C bond cleavage through nitrogen-centered radicals (Scheme 1a).

Scheme 5
scheme 5

Formation of z• radical from x• radical by radical-induced cleavage

Full size image
Figure 8
figure 8

Rate constants for isocyanic acid loss from x3• with an Xxx-Ala-Ala-NH2 sequence, where Xxx is Gly, Ala, or Asn. The horizontal dashed line at k = 107 corresponds to dissociations occurring on the MALDI-ISD timescale of 100 ns

Full size image

In addition to x fragments, w and y fragments were observed in the spectrum shown in Figure S4. Since MALDI-ISD experiments with a reducing matrix also produced w and y fragments [11], the origin of these fragments might not be hydrogen-deficient radicals; however, the formation mechanism of these fragments was not ascertained. It should be noted that the spectrum in Figure S4 shows an x10 fragment, which is formed by a cleavage of on the C-terminal side of Gly residue, with moderate intensity, even though the abundance of a fragments originating from the Cα–C bond cleavage on the C-terminal side of the Gly residue is known to be low. Nagoshi et al. [30] concluded that the low abundance of the corresponding a fragment is due to the low efficiency of Cα–C bond cleavage, arising from the lack of β-carbon atoms. However, the counterpart x fragment is present in the MALDI-ISD mass spectrum, suggesting that the C-terminal side of the Gly residue might be cleaved during MALDI-ISD, as is the case for other Cα–C bonds. One explanation for the low abundance of a• radical originating from the cleavage of the C-terminal side of the Gly residue is that a further radical dissociation pathway is due to the low stability. This point will need future investigation.

Conclusion

Although nitrogen-centered and β-carbon–centered hydrogen-deficient peptide radicals are generally considered to be intermediates in this process when using an oxidizing matrix, the results of the present joint experimental and computational study strongly suggest that peptide Cα–C bond cleavage occurs through nitrogen-centered radical formation. These results indicate that the oxidizing matrix preferentially abstracts a hydrogen atom from the amide nitrogen in peptide bond with subsequent radical-induced Cα–C bond cleavage, leading to the generation of an a•/x fragment pair. Then, the a• radical fragment either reacts with a matrix molecule or loses its side chain, leading to the a or d fragments. The yield of d fragments is dependent on the dissociation rate constant of the corresponding a• radical. In contrast, the matrix abstracts a hydrogen atom from either the amide nitrogen atom or the β-carbon atom in a• radical, producing an a fragment.